Prediction of Heat-Treated Spruce Wood Surface Roughness with Artificial Neural Network and Random Forest Algorithm

Abstract

AbstractWood material has a wide range of uses due to its many positive properties. In addition to its positive features, it also has negative features that limit the usage area of wooden material. One of the commonly used methods to minimize these negative properties is heat treatment application. In the study, the surface roughness values of Spruce (Picea abies) samples heat treated with ThermoWood method were investigated. Surface roughness measurements were carried out in the radial and tangential directions with the Mitutoyo SJ-201M tactile surface roughness tester. Then, the contact angle values ​​of the samples in the tangential and radial direction were determined. TS 4084 standard was used to determine the swelling and shrinkage amounts of the samples whose contact angle values ​​were determined. Surface roughness values of the samples were estimated by artificial neural network (ANN) and random forest algorithm. In the estimation of contact angle with random forest algorithm and ANN method, swelling and shrinkage amounts were entered as input. In the study, it has been determined that the predictions made in the radial direction with artificial neural networks give the most accurate results. In predictions made in the radial direction with artificial neural networks, R2 = 0.98 and RMSE = 0.03. In the radial study conducted with Random Forest Algorithm, R2 = 0.96 and RMSE = 0.11. As a result, it has been determined that the surface roughness of a wood material can be estimated by ANN and Random forest algorithm.KeywordsHeat treatmentSpruceSurface roughnessArtificial neural networkRandom forest algorithm